Shaped Strong Transition Model

This page first posted 4 April 2015

In time for the upcoming general election, Electoral Calculus has upgraded its prediction model. The new model builds on the success of the previous models, but has a number of new features:

The previous model was the Strong Transition Model. It took essentially a uniform national picture and applied to each seat, but with two extra sophisticated features. For declining parties it behaves multiplicatively - in other words, if a party loses a quarter of its support nationally, then it loses a quarter of support in each seat. But gaining parties gain additively taking a set share of votes "released" by the declining parties. Secondly, it divided voters into "weak" and "strong" supporters - and the weak supporters drift away before any strong ones do.

This model works quite well, but the upcoming election has shown up two weaknesses. The first of these problems is caused by the rise of the two new minor parties: UKIP and the Greens. These parties polled only around 1%-3% in the 2010, but are now polling around 6%-16%. This has been mostly at the expense of the Liberal Democrats, who have declined from 24% to 8%. The Strong Transition Model moves Lib Dems votes towards UKIP and the Greens. But this gives, say, UKIP fairly equal strength across the whole country. And it makes UKIP particularly strong in areas of previous Lib Dem strength, such as Scotland and Wales. But this is not accurate, because these are not really areas of UKIP strength, such as Essex and Lincolnshire.

UKIP areas of strength and weakness

The new model performs a very granular geographic treatment, giving a better prediction of UKIP in each seat. This means there is also more variation in UKIP votes between seats, with the 25%-percentile of vote strength decreasing from 14% to 12%, and the 75%-percentile increasing from 18% to 20%. This means that the semi-inter quartile range (gap between the 75% strongest vote share and the 25% strongest vote share) has approximately doubled.

The top ten UKIP seats for the old and new models are shown below:

Seat (old model)RegionPred UKIP
votes
BuckinghamEssex27%
Stoke-on-Trent CentralSevern25%
BromsgroveSevern23%
Great GrimsbyHumberside23%
Hull EastHumberside23%
MiddlesbroughNorth23%
Boston and SkegnessHumberside23%
Wentworth and DearneHumberside22%
Stoke-on-Trent NorthSevern22%
Cambridgeshire North WestEast Anglia22%
Seat (new model)RegionPred UKIP
votes
ClactonEssex41%
Rochester and StroodSouth East30%
Stoke-on-Trent CentralSevern29%
Boston and SkegnessHumberside28%
BuckinghamEssex27%
Wentworth and DearneHumberside26%
Great GrimsbyHumberside26%
South Holland and The DeepingsHumberside26%
Cambridgeshire North EastEast Anglia26%
Louth and HorncastleHumberside26%

We see that the greater range of the new model means that the top 10 seats have more predicted support than the equivalent top 10 seats under the old model. This can also be seen in the following table of average UKIP predicted votes over the various regions, under both the old and new models.

RegionOld
Model
New
Model
Humberside19%22%
Essex18%20%
South West18%20%
South East17%19%
Severn18%19%
East Anglia18%19%
North17%18%
East Midlands17%17%
South17%17%
West Midlands17%17%
Greater Manchester18%16%
Yorkshire16%16%
Lancashire16%15%
West16%15%
Wales15%15%
South London15%11%
North London14%8%
East Scotland5%5%
West Scotland4%4%

Humberside, Essex and the south of England are more UKIP than an even geographic assumption implies. On the other hand, London is much less UKIP supporting, apart from London boroughs close to Essex.

Green areas of strength and weakness

Let's look at the Greens to see what geographic variation their support has. Similarly the UKIP, the semi-inter quartile range approximately doubles in width from 1.7% (old range from 5.1% to 6.8%) to 3.1% (new range from 4.1% to 7.2%). This shows that the new model allows for more geographic variation than the straight STM.

The top ten Green seats for the old and new models are shown below:

Seat (old model)RegionPred Green
votes
Brighton PavilionSouth East35%
Norwich SouthEast Anglia20%
CambridgeEast Anglia13%
Lewisham DeptfordSouth London12%
Leeds WestYorkshire11%
Lancaster and FleetwoodLancashire10%
Hackney North and Stoke NewingtonNorth London10%
HoveSouth East10%
Brighton KemptownSouth East10%
HuddersfieldYorkshire10%
Seat (new model)RegionPred Green
votes
Brighton PavilionSouth East38%
Norwich SouthEast Anglia24%
CambridgeEast Anglia18%
Hackney North and Stoke NewingtonNorth London16%
Lewisham DeptfordSouth London16%
Oxford EastWest16%
Bristol WestWest15%
Hackney South and ShoreditchNorth London15%
Bristol SouthWest15%
Bristol EastWest14%

The strongest Green seats are now more strong than previous estimates, with Green support generally about 5% higher. The strong seats are more concentrated in the south of England than before. We can see the regional breakdown in the table below.

RegionOld
Model
New
Model
West6.5%9.0%
South West6.1%7.8%
South London6.9%7.6%
North London6.6%7.5%
South East7.1%7.2%
East Anglia7.2%7.1%
Yorkshire6.7%7.1%
South6.0%6.6%
Greater Manchester6.3%6.1%
Essex6.1%5.8%
Humberside6.3%5.6%
Lancashire5.5%5.5%
East Midlands6.0%5.2%
North6.1%5.0%
Severn6.0%4.8%
West Midlands5.7%4.0%
East Scotland3.2%3.4%
Wales5.3%3.4%
West Scotland2.7%2.8%

Strong regions for the Greens are in the West/South West, London and the South East. Weaker regions for the Greens are Scotland, Wales and the north of England.


Return to Psephology page, or home page.